Exp Econ (2009) 12: 26–41

DOI 10.1007/s10683-007-9188-7

Group polarization in the team dictator game

reconsidered

Wolfgang J. Luhan · Martin G. Kocher ·

Matthias Sutter

Received: 8 August 2006 / Revised: 19 July 2007 / Accepted: 26 July 2007 /

Published online: 2 November 2007
© Economic Science Association 2007

Abstract While most papers on team decision-making find that teams behave more
selfishly, less trustingly and less altruistically than individuals, Cason and Mui (1997)
report that teams are more altruistic than individuals in a dictator game. Using a
within-subjects design we re-examine group polarization by letting subjects make
individual as well as team decisions in an experimental dictator game. In our ex-
periment teams are more selfish than individuals, and the most selfish team member
has the strongest influence on team decisions. Various explanations for the different
findings in Cason and Mui (1997) and in our paper are discussed.

Keywords Experiment · Dictator game · Team behavior · Social preferences

JEL Classification C72 · C91 · C92 · D70

W.J. Luhan
Center for Social Science Methodology, University of Oldenburg, Ammerlaender Heerstrasse
114-118, 26129 Oldenburg, Germany

M.G. Kocher ()

Department of Economics, CREED, University of Amsterdam, Roetersstraat 11, 1018 WB
Amsterdam, The Netherlands
e-mail: martin.kocher@uibk.ac.at

M.G. Kocher · M. Sutter

Department of Public Finance, University of Innsbruck, Universitaetsstrasse 15/4, 6020 Innsbruck,
Austria

M. Sutter
Department of Economics, University of Göteborg, P.O. Box 640, 40530 Göteborg, Sweden
Group polarization in the team dictator game reconsidered 27

1 Introduction

Team decision-making has attracted more and more interest among economists in
recent years. Although the standard models of economic theory largely neglect the
type of the decision-maker by simply assuming individual decision-making, many
economic decisions are made by teams such as families, company boards, manage-
ment teams, committees, or central bank boards. One of the key requirements of team
decision-making is the need to aggregate individual preferences into one single team
decision.
In this paper we will examine the aggregation of social preferences in an exper-
imental team dictator game. This simple game isolates pure distributional concerns
from any strategic considerations that shape bargaining behavior, for instance, in the
well-known ultimatum game (Güth et al. 1982).1 Hence, using the dictator game al-
lows us to investigate how individual preferences with respect to the allocation of
money between a dictator and a (powerless) recipient are transformed into a team
decision.
In general, the experimental literature on unitary team2 behavior in simple bar-
gaining games is rather coherent in its finding that team decisions are typically
closer to standard game theoretic predictions than individual decisions, that teams
seem to be more strongly motivated by payoff maximization and that their de-
cisions are more selfish. For example, teams have been found to send and ac-
cept smaller transfers in the ultimatum game (Bornstein and Yaniv 1998), send
(Kugler et al. 2007) or return (Cox 2002) smaller amounts in the trust game,3
choose lower efforts as second movers in a gift-exchange game (Kocher and Sut-
ter 2007), play more strategically and more often according to theoretical predic-
tions in signaling games (Cooper and Kagel 2005), exit the centipede game at
earlier stages (Bornstein et al. 2004), make more rational decisions in intellec-
tive tasks, such as the Wason selection task (Maciejovsky and Budescu 2007) or a
non-interactive common-pool-resource game (Gillet et al. 2007), and gain a higher
payoff/risk ratio in investment games (Rockenbach et al. 2007) or higher payoffs
in lottery choices (Sutter 2007). In auctions, teams display much more competi-
tive bidding behavior leading to higher rates of overbidding (Cox and Hayne 2006;
Sutter et al. 2006).
In contrast to this large body of evidence, Cason and Mui (1997) find teams to
be more altruistic and other-regarding than individuals. In their experimental dictator

1 Forsythe et al. (1994) used the dictator game in order to study which share of the transfers observed in
the ultimatum game can be explained by pure altruistic motivations. By removing the responder’s option
to reject the proposer’s offer, any strategic reason for generous offers in the dictator game is eliminated,
thus leaving only social preferences such as altruism, warm glow or inequality aversion as possible expla-
nations.
2 With “unitary” teams we denote teams that do not face any internal conflict in terms of payoff and that
must make a single team decision.
3 The findings of Cox (2002) and Kugler et al. (2007) are not completely consistent. Cox (2002) finds no
differences between teams and individuals in the roles of senders but observes that teams return signifi-
cantly less in the role of responders. Kugler et al. (2007) report the opposite with teams sending smaller
amounts than individuals, whereas returns of teams and individuals do not differ.
28 W.J. Luhan et al.

game, each subject has to make two decisions on the allocation of five dollars, once
individually and once in a two-person team. Cason and Mui (1997) show that team
decisions are driven by the more altruistic team member. They conclude that “this
makes team choices more other-regarding than individual choices for these teams,
although the difference is modest” (p. 1480). More precisely, individuals transfer on
average 26.09% of their endowment to the receiver, compared to the 27.61% of teams.
Cason and Mui (1997) explain this effect by referring to “Social Comparison Theory”
(SCT). SCT states that subjects have a tendency to appear and present themselves
in teams in a way that is deemed socially desirable. After having observed other
subjects’ behavior or choices or after having discussed the possible options in a game,
the own behavior is modified to appear more in line with the perceived social norms.
If other-regarding behavior is socially desirable, then team decisions can be expected
to be more other-regarding than individual decisions.4
Though (or perhaps because) the results of Cason and Mui (1997) do not fit the
overall picture emerging from team decision experiments, they have been frequently
cited and are widely accepted as a notable exception in the evidence on team decision-
making (see, for example, Bosman and van Winden 2002; Andreoni and Petrie 2004;
Bornstein et al. 2004; Cooper and Kagel 2005; Kocher and Sutter 2005; Dufwenberg
and Muren 2006). Given the prominence that the results of Cason and Mui (1997)
have achieved, and since we consider replications of results important in experimen-
tal economics, this paper examines the robustness of Cason and Mui’s (1997) find-
ings. Moreover, it adds some interesting new insights on the decision-making process
within teams.
We are not aware of any other attempt to replicate the findings of Cason and Mui
(1997).5 A somewhat related paper is by Dufwenberg and Muren (2006) who study
the influence of the gender composition of teams on giving in a dictator game. How-
ever, they do not compare individual to team decisions, which is the main focus of
our paper.
The paper is organized in the following way: We describe our experimental de-
sign and procedure in Sect. 2. Section 3 presents the experimental results. Section 4
discusses our findings and concludes.

2 Experimental design and procedure

The dictator game was first introduced by Kahnemann et al. (1986) in an empirical
study on fairness in the market place. In this game, a dictator decides on the distrib-

4 Cason and Mui (1997) reject “Persuasive Argument Theory” (PAT) as a possible explanation for their
data. The bottom-line of PAT is that team discussion is able to shift choices in favor of the pre-discussion
or initial tendency by a higher attentiveness towards more persuasive arguments in favor of one’s initial
position. Both theories, SCT and PAT, have their roots in the psychological discussion of “group polar-
ization” and choice shifts (see e.g., Stoner 1961; Teger and Pruitt 1967; Moscovici and Zavalloni 1969;
Davis 1992; Kerr et al. 1996; Levine and Moreland 1998). A general model of choice shifts (for choices
under uncertainty) can be found in Eliaz et al. (2006).
5 Note, of course, that second movers in a trust game are basically in the situation of a dictator, as they
can determine the final distribution of money between the first and the second mover. However, players’
motivations in the trust game can involve reciprocity, which constitutes a clear difference to the dictator
game.
Group polarization in the team dictator game reconsidered 29

ution of a given stake between himself and a second party, the receiver. The second
party does not have any other option than to take whatever the dictator allocates to
her.
In our experiment participants were randomly assigned to the role of either dictator
or receiver.6 These roles were fixed throughout the experiment. Subjects in the role
of dictator received 5 € as their endowment in each of three stages of the experiment.
The following describes the three stages of our main experimental treatment, denoted
as TEAM-treatment.
• Stage S1: Each subject in the role of the dictator decides individually on the trans-
fer T to an individual receiver, where T ∈ {0.0, 0.1, 0.2, . . . , 4.9, 5.0}.
• Stage S2: Each subject in the role of the dictator is a member of a three-person
team. Each team has to make a single decision on the transfer T to a receiver team
that also consists of three subjects. The per-capita incentives are kept identical in
all stages, such that in stage S2 each team member in the role of the dictator earns
5 − T €, and each member in the receiver-team earns T €.
• Stage S3: Each subject in the role of the dictator has to make a second individual
decision under the same rules as in stage S1.
In order to separate the effects of team decision-making from the possible learning
effects across the three stages in the TEAM-treatment, we have set up a CONTROL-
treatment in which the second stage S2 is also played individually. Hence, in the
CONTROL-treatment subjects make three individual decisions in a row. Compar-
ing the data from stage S2 in both treatments provides a (between-subjects) way of
discriminating order effects (through repetitions of the same task) and team vs. indi-
vidual treatment effects.7
In both treatments, participants were informed that the experiment consisted of
three consecutive stages and that decisions in each stage were completely indepen-
dent and would not have any impact on subsequent stages. There was no feedback
of any kind before the end of the experiment, and participants were only informed
about the rules of the next stage after all decisions in a respective stage had been
made. Given this procedure we regard it as rather unlikely that any decision could
have been influenced by an (at that stage unknown) later decision. We used a pub-
licly announced perfect stranger design throughout the three stages, meaning that no
dictator was ever paired with a particular receiver more than once.
An important feature of our design is the assignment of individuals into teams in
our TEAM-treatment. First of all, we opted for teams of three subjects each in order
to address Cason and Mui’s (1997) suggested extension of checking whether their
empirical findings generalize when the size of teams gets larger (p. 1480). Extending
the team size from two members to three members may have important implications
for the decision-making procedure within a team. With two members, it is most likely

6 Since we expected that receivers would earn little money, they were invited to participate in an individual
decision-making experiment. This second experiment was announced and conducted after the dictator
experiment had been finished.
7 We are grateful to an anonymous referee and James Cox for suggesting this CONTROL-treatment as a
robustness check.
30 W.J. Luhan et al.

that unanimous (compromise) decisions will be taken (unless one member voluntarily
surrenders to the decision of the other member). With three members, applying an im-
plicit majority rule becomes an option. This latter feature of three-person teams might
have an impact on resulting team decisions. However, only very little is known about
the influence of team size on team decisions, in particular when social preferences
are involved. We are only aware of one study by Wildschut et al. (2001), who have
not found any differences between the cooperation levels of two- and three-person
teams in a prisoner’s dilemma game.
Second, we have used the individual stage S1-choices as an instrument to set up
heterogeneous teams in a systematic way. Based on their stage S1-decision we have
classified subjects into terciles. Those in the lowest tercile are the “most selfish” sub-
jects, those in the highest tercile the “most other-regarding”, and those in the middle
the “moderate” subjects. Within each tercile, subjects are ranked in ascending order
according to their individual transfer in stage S1. Each team then consists of the three
members with the same rank in each of the three terciles. This procedure guarantees
that all teams are very similar with respect to the within-team heterogeneity concern-
ing individual transfers from stage S1. It will allow us to track the influence of the
most selfish, the moderate, or the most other-regarding member on the final team
decision in a straightforward manner.
In order to gain further insights into the process of aggregating individual prefer-
ences and into the structure of team discussions, the communication within each team
was possible exclusively via an electronic chat program with protocols being stored
for later analysis.8 Subjects were requested to agree on a joint (unanimous) decision
within ten minutes. As soon as a team had reached an agreement, each member had
to type in this decision on his own screen. Note that we did not specify how teams
had to arrive at a joint decision. Even though the inputs of all three members had to
be identical to be valid—otherwise the team would not have been paid for this stage
(which never happened, actually)9 —some teams argued in the chat that a simple ma-
jority in case of no unanimous agreement would be sufficient and the third member
would have to abide by the majority in such a case. Of course, there was no way to
enforce such a within-team agreement.
At the end of the experiment we gathered data on subjects’ gender, allowing us
to examine the influence of gender and gender composition on individual and team
decisions. The experiment was computerized (using z-Tree; Fischbacher 2007) with
220 students from the University of Innsbruck from various fields of study. 180 stu-
dents participated in the TEAM-treatment, yielding 180 observations on individual
decision-making (90 in stage S1 and 90 in stage S3) and 30 observations from teams
along with concomitant chat-protocols (from stage S2). 40 students participated in the
CONTROL-treatment, yielding 20 decisions of dictators in each of the three stages.

8 The content of messages within a team was not restricted, except for forbidding abusive language and
revealing one’s identity (through seat number, name, gender, age, courses taken, etc.). See the experimental
instructions that can be obtained from the web page of the journal for details.
9 The instructions specified that in such a case the transfer to the paired team would have been determined
randomly.
Group polarization in the team dictator game reconsidered 31

Table 1 Transfers in the dictator game

A. TEAM-treatment Stage S1 Stage S2 Stage S3

1st individual transfer Team transfer 2nd individual transfer
(N = 90) (N = 30) (N = 90)

Average transfer T 0.94a,b 0.54a,d 0.66b,c

Standard deviation 1.07 0.56 0.85
Mode (= Minimum) 0.00 0.00 0.00
Relative frequency of 34.4% 40.0% 46.7%
Mode (= Minimum)
Maximum 5.0 1.5 2.5
Relative frequency of 1.1% 13.3% 8.9%
Maximum
B. CONTROL-treatment 1st individual transfer 2nd individual transfer 3rd individual transfer
(N = 20) (N = 20) (N = 20)
Average transfer T 1.27 1.17d 1.25c
Standard deviation 1.25 1.17 1.11
Mode (= Maximum) 2.5 0/2.5 2.5
Relative frequency of 45% 35% 35%
Mode
Minimum 0.00 0.00 0.00
Relative frequency of 40% 35% 25%
Minimum
a Significantly different at p < 0.01

b Significantly different at p < 0.01

c Significantly different at p < 0.05

d Significantly different at p = 0.069

3 Results

Table 1 shows the main indicators with respect to the chosen transfers. Panel A of the
table refers to the TEAM-treatment, and panel B presents data from the CONTROL-
treatment.
In stage S1 of the TEAM-treatment, individuals send on average 0.94 € to the
receiver, i.e. 19% of the endowment.10 The modal transfer (of about one third of sub-
jects) is zero. The transfers of teams in stage S2 are significantly smaller than those of
individuals in stage S1 (p < 0.01; Wilcoxon signed ranks test; N = 30). Teams trans-
fer only 0.54 € or about 11% of their endowment. The maximum transfer is 1.5 €
out of 5 €, and 12 out of 30 teams transfer zero. It turns out that the experience of
team decision-making even carries over to individual decisions in stage S3. On aver-
age, the transfers in S3 are 0.66 € (or about 13%), and they are significantly smaller

10 This average is well within the range of transfers (10% to 23%) reported in the survey on experimental
dictator games by Camerer (2003).
32 W.J. Luhan et al.

Fig. 1 Changes of individual decisions contingent on team decision (multiple observations indicated)

than individual transfers in S1 (p < 0.01; Wilcoxon signed ranks test; N = 90), but
not significantly different from the team decisions in stage S2 (p > 0.5).
In the CONTROL-treatment, we find that average transfers stay basically stable
across all three stages. It is important to note that transfers in the first stage S1 are
not significantly different between both treatments (0.94 € vs. 1.27 €; p > 0.3;
Mann–Whitney U-test; N = 110). However, transfers in the second stage S2 are sig-
nificantly different between teams (in the TEAM-treatment) and individuals (in the
CONTROL-treatment), indicating that team decision-making drives down transfers
and not the repetition of the task (0.54 € vs. 1.17 €; p = 0.069; Mann–Whitney
U-test; N = 50). Transfers in the third stage S3 are again significantly lower in the
TEAM-treatment than in the CONTROL-treatment (0.66 € vs. 1.25 €; p < 0.05;
Mann–Whitney U-test; N = 110). This latter result confirms the influence of the
experience of team decision-making on the final individual decision (contrary to a
possible learning effect across stages).
Figure 1 sheds more light on the influence of team decision-making on subsequent
individual decisions in the TEAM-treatment. It shows a subject’s individual decision
in stage S1 on the horizontal axis and the individual decision in stage S3 on the
vertical axis. Observations below the diagonal indicate a decrease of transfers from
S1 to S3. Individuals are classified according to the relationship between the team’s
decision in stage S2 and their first individual decision in stage S1. Note that when
the team decision is larger than (see “+” in the figure) or equal to the first individual
decision in stage S1 (see “◦”), then there is no a systematic effect of the team decision
on the subsequent individual decision in stage S3. However, if the team decision is
smaller than the S1-decision (see “–”) individuals reduce their transfer from S1 to S3
Group polarization in the team dictator game reconsidered 33

Table 2 OLS and Tobit estimates for team transfers in the TEAM-treatment

Dependent variable: team transfer OLS Tobit

Coefficient Robust standard Coefficient Standard
errors errors

Transfer of “most selfish” member 1.037a 0.46 1.554c 0.912

in stage S1
Transfer of “moderate” member in 0.131 0.19 0.213 0.247
stage S1
Transfer of “most other-regarding” −0.094 0.08 −0.215 0.209
member in stage S1
Constant 0.615b 0.20 0.556 0.445

Observations 30 30
R2 /Log likelihood 0.08 −31.3574
a p < 0.05

b p < 0.01
c p = 0.11

in almost all cases, indicated by the “–” signs below the 45-degree line. The latter
result implies a strong influence of team decision-making on individual decisions if
the team decision is more selfish than the initial individual decision, rendering the
subsequent individual decision also more selfish.
Next we examine the process of team decision-making. We have estimated a linear
as well as a Tobit model explaining a team’s transfer in S2 as a function of its mem-
bers’ individual transfers in the preceding stage S1. Given that both models yield
qualitatively similar results, and due to the low number of censored observations, we
will only discuss the results from the OLS estimation (see Table 2). The “most self-
ish” team member (i.e. the team member with the lowest transfer in S1) has by far
the largest and, in particular, the only significant impact on the team transfer.11 The
lower the “most selfish“ member’s transfer in S1, the lower is the team transfer in
S2. This finding is in contrast to the evidence from Cason and Mui (1997), who have
found that the team transfer is driven by the most other-regarding member.
In order to gain further insights into the driving forces behind a team’s decision
we resort to the contents of the chat protocols. On average, teams exchanged 22.3
messages (125 words) before reaching an agreement.12 However, the variance was
rather large, with the shortest chat including only five messages with 18 words and the
longest one including 58 messages with 353 words (see the Appendix for a translation
of both). In Table 3 we list the arguments that have been voiced in at least two separate
teams.

11 Cox (2002) reports the same pattern for the second mover teams in his team trust game.
12 Neither the number of exchanged messages nor the number of words has any statistically significant
impact on team transfers in stage S2.
34 W.J. Luhan et al.

Table 3 Main arguments in the electronic team chat in the TEAM-treatment

Argument Number of teams where Relative frequency of Average transfers

argument is discussed teams with argument

Keep more for ourselves 22 73.3% 0.51

Fairness and . . .
. . . smaller transfers 11 36.7% 0.52
. . . higher transfers 6 20.0% 0.70
Rationality 4 13.3% 0.25
Economic expertise 2 6.7% 0.00

Table 4 Main arguments and number of words in the chat of the TEAM-treatment

Argument More for Fairness and Fairness and Rationality Economic Number
Member ourselves smaller transfers higher transfers expertise of words

“Most selfish” 37.93% 56.25% 37.93% 50.00% 50.00% 33.53%

“Moderate” 31.03% 18.75% 30.77% 25.00% 50.00% 32.01%
“Most other- 31.03% 25.00% 30.77% 25.00% 0.00% 34.46%
regarding”

The argument of sending small transfers in order to “keep more for ourselves” is
voiced most often, namely in 22 out of 30 teams, and these teams send on average
a transfer of 0.51 €. This argument appeals to group identity by promoting a better
outcome for the members of the team. It is never associated with the blunt claim that
the team should behave selfishly.
The notion of “fairness” is mentioned in 17 out of 30 teams. It is used, however, in
two markedly different contexts. There are 11 teams in which remarks on fairness are
actually used to support less generous transfers, and a typical argument is that “they
[the recipients] can’t expect us to send a fair share since they wouldn’t do it either”
(Team T5). These teams send on average 0.52 €. In 6 teams the notion of fairness is
put forward to induce more generous transfers or to prevent relatively small transfers.
One example is that “. . . it’s not fair to send only 20 cents, but we should send 1
euro” (Team T18). When fairness is mentioned in this context, the average transfer is
0.70 €.
In four teams one member mentions that it would be “rational” to send a specific
amount, without however defining rationality. Two subjects in different teams pro-
posed zero transfers and justified their proposal based on their “economic expertise”.
In sum, the arguments put forward seem to support the observation that smaller—
rather than larger—transfers were the predominant social norm, often supported by
arguments linked to payoff maximization or rational decision-making, as indicated in
Table 3. Under these circumstances both PAT as well as SCT would, in fact, predict
smaller transfers of teams than of individuals, which is what we actually observe.
The results from Tables 2 and 3 raise the question of why the most selfish team
member has such a strong impact on the team’s decision. Table 4 breaks down the
Group polarization in the team dictator game reconsidered 35

Table 5 Decision rules, team

composition and transfers Decision rule Other-regarding (OR) Self-regarding (SR)
N Av. transfer N Av. transfer

Unanimity 2 0.50 5 0.70

Majority 3 0.00 4 0.25
Compromise 8 0.89 7 0.31

main arguments put forward in the electronic chat by the type of member (most self-
ish, moderate or most other-regarding from stage S1). The first thing to notice from
the right-most column is that the relative frequency of words written in the chat is
basically the same over the three types. Hence, the most selfish members do not ar-
gue more actively. On average, however, we notice that the most selfish members of
a team invoke arguments of fairness in connection with smaller transfers, rationality
and economic expertise more often than the other members, even though the differ-
ences are not statistically significant due to the rather low number of observations. In
sum, the evidence in Table 4 suggests that the mere presence of a selfish team mem-
ber or the expression of selfish ideas (promoting small transfers as being fair; see also
Table 3) induces smaller transfers in many teams.
As noted in Sect. 2 above, teams with three members may resort to majority-rule
decision-making in the event that there is not a unanimous agreement on a particu-
lar transfer. As a final analysis of the team decision-making process, we summarize
in Table 5 the decision rules actually applied and the associated transfers. A unani-
mous decision was identified only if all team members independently suggested the
same transfer or if the first stated option was accepted by all three members without
discussing any alternatives (see “Unanimity”). If there was at least one alternative
proposal that was overruled, or if at least one member had to be convinced of an
alternative transfer, this team was classified as using a “Majority” rule. A “Compro-
mise” was reached if all three members agreed to a transfer that was different from
their first proposal. In Table 5 we also classify teams with respect to their members’
preferences. If the “moderate” member’s individual transfer in stage S1 was below
the overall median13 of 0.5 €, a team was classified as self-regarding (SR); if it was
above the overall median it was considered as an other-regarding (OR) team.
We expected to find self-regarding teams to send lower transfers than other-
regarding teams. Yet, Table 5 does not fulfill this expectation, since the lowest av-
erage transfers were actually sent by other-regarding teams that used a majority rule.
Overall, we do not find any significant differences between the entries in Table 5
(Kruskal–Wallis-Test; p > 0.1). Hence, the type of decision-making rule applied in a
team does not seem to have a systematic influence on the chosen transfers, although
one has to bear in mind that the number of observations in each category of Table 5
is relatively small.14

13 Following Cason and Mui (1997), we define the median of all individual transfers in stage S1 of the
TEAM-treatment as our “neutral point”.
14 The same holds for gender composition of teams. We do not find any significant differences in transfers
between teams with at least two women (0.56 €) and teams with at least two men (0.53 €). In the indi-
36 W.J. Luhan et al.

4 Conclusion

Our experimental results indicate that teams choose significantly smaller transfers
than individuals in a dictator game and that individuals shift their transfers towards
the team transfer when asked to make a second decision after team interaction. Even
though the differences between individual and team decisions are modest in absolute
terms (amounting to 40 Eurocents between stages S1 and S2), the difference is quite
large in relative terms (43%). Hence, team decisions appear to be noticeably more
selfish than individual ones.
This main result of our paper is in line with the large majority of experimental
papers that have shown that team decisions are more selfish and competitive, less
trusting and less altruistic than individual decisions (see, e.g., Bornstein and Yaniv
1998; Bornstein et al. 2004; Cooper and Kagel 2005; Cox 2002; Kocher and Sutter
2005; Kugler et al. 2007).
Our data are in contrast to Cason and Mui’s (1997) results, though. In the follow-
ing we would like to discuss several possible explanations for the different findings.
First, Cason and Mui (1997) have controlled for the order of decision-making by
having two treatments, one with the sequence “individual-team” decision-making,
and another with the reverse order “team-individual”. They find teams to be more al-
truistic in both treatments. We have not been able to control for the order of decision-
making in the way Cason and Mui (1997) did, because our procedure for setting up
teams required an individual decision prior to the team decision-making stage and,
therefore, rules out a team decision as the first stage. Yet, we regard it as unlikely that
order effects might have caused the differences, because (1) Cason and Mui (1997)
did not observe an order effect, and (2) our treatment variation (TEAM vs. CON-
TROL) produced no significantly different transfers in the first stage, irrespective of
whether there was a team or an individual decision in the second stage, but it yielded
significantly smaller transfers of teams than of individuals in the second stage.
Second, Cason and Mui (1997) have implemented two-person teams, whereas
we have set up teams with three members. So far, there is little systematic vari-
ation of the size of teams in economic experiments and little evidence regard-
ing its influence on team decisions where social preferences are important.15 Yet,

vidual decision-making stages of the TEAM-treatment—as in the CONTROL-treatment—we also do not

find any differences in transfers of men or women (detailed figures are available upon request). Recall that
Dufwenberg and Muren (2006) report an impact of gender. Yet, in their design the gender (composition) of
teams was common knowledge, whereas it was not known in our design (the revelation of one’s gender was
forbidden in the chat). The seeming contradiction can be reconciled by the finding of Bolton and Katok
(1995), who have observed that those experiments that reveal subjects’ gender typically find significant
gender effects, while studies without such information usually fail to report significant effects (see also
Andreoni and Vesterlund 2001, or Cox and Deck 2006, for a systematic treatment of gender differences in
bargaining).
15 In the context of a beauty-contest game Sutter (2005) has shown that larger teams (of four subjects)
win the game more often than smaller teams (of two subjects) or individuals. The superiority of (larger)
teams might be explained by the fact that team discussion forces team members to approach a problem
more systematically and that team discussion provides ample opportunity for error-checking (see Davis
and Harless 1996, for an account of these advantages of team decision-making). Yet, these explanations
apply much more to intellective tasks (such as a beauty-contest game or monopoly pricing) than to tasks
where social preferences play a major role.
Group polarization in the team dictator game reconsidered 37

the scarce evidence does not support the conjecture that three-person teams act
differently than two-person teams. Wildschut et al. (2001) do not report differ-
ences between two-person and three-person teams in a prisoner’s dilemma game.
Comparing across different studies, teams have been found to be more strategic
(e.g., Cooper and Kagel 2005) and less altruistic (e.g., Bornstein and Yaniv 1998;
Cox 2002) with two, three or five members. Our evidence from Table 5 also implies
that the decision-making rule (unanimity versus majority rule, with the latter not be-
ing possible in two-person teams) does not cause significant differences. In sum, it
seems therefore unlikely that the different size of teams may have caused the differ-
ences between our study and the one by Cason and Mui (1997).
Third, in the experiment of Cason and Mui (1997) “every team was called to the
front of the room (by identification numbers) and excused to the hallway to discuss
their decision and fill out the form in private” (p. 1471). This procedure entails two
further, possibly important sources for the contrasting evidence. On the one hand,
calling team members to the front of the room identifies the members of a team.
Even though their decisions have remained anonymous in the experiment, the mere
identification may have caused the higher transfers of teams. In fact, the evidence
from Bohnet and Frey (1999) shows that transfers in the dictator game increase with
identification of subjects (without revealing decisions). On the other hand, the teams
in Cason and Mui’s (1997) experiment used face-to-face discussion, whereas in our
experiment team members interacted only via an electronic chat. Both approaches
have their merits, but need not be neutral with respect to their influence on final deci-
sions. The face-to-face design captures a richer contextual field of personal encoun-
ters when teams make decisions. The electronic chat-approach provides a more con-
trolled environment by preserving anonymity and avoiding confounding factors such
as, for instance, personal sympathy or antipathy or prior acquaintance of team mem-
bers unknown to the experimenter. The differences between both approaches might
induce different degrees of social distance (within and across teams), which could
trigger the different results. In fact, Kocher and Sutter (2007) have found first mover-
teams in a gift-exchange game to act differently, depending upon whether team mem-
bers are requested to reach a decision via face-to-face communication or via voting
electronically on different proposals. When using the electronic form of communica-
tion, teams make more selfish decisions.16 The same mechanism could be the driving
force behind the different findings in Cason and Mui (1997) and the current paper,
meaning that there is no fundamental inconsistency between the two contributions.
Obviously, studying the effects of different communication media on team decision-
making seems to be an important avenue for future research in order to enhance our
understanding of team decision-making.
Acknowledgements We would like to thank Tim Cason, Jim Cox (as guest editor for this paper) and
three anonymous referees for excellent comments. Olga Mellizo did a great job in copy-editing. Finan-
cial support from the Austrian Science Foundation (FWF Project P16617), the Tiroler Wissenschaftsfonds
and the Center for Experimental Economics at the University of Innsbruck (sponsored by Raiffeisen-
Landesbank Tirol) is gratefully acknowledged.

16 A somewhat related result can be found in Güth et al. (2007). They show in a large-scale newspaper
experiment that (individual) decisions in a three-person ultimatum game are more selfish if submitted via
Internet than via the more personal form of sending a letter.
38 W.J. Luhan et al.

Appendix Two examples of the chat conversation

The initial number in each line indicates a particular team member (1 to 3). Italic text
at the end of a chat refers to the individual decisions in stages S1 and S3.

1. The most extensive discussion

1: Obvious decision, don’t you think?

3: Sure, all for ourselves.
1: I am no good Samaritan.
3. Transfer: 0.
3: Right, neither am I.
1: We are not responsible for b, are we?
1: Therefore 0.
1: OK. I’ll enter it.
3: No. 2, do you share our opinion?
2: I think we should be fair.
3: I don’t. Nobody has ever been fair to me.
1: 0 is fair.
2: Wait a second don’t enter anything.
3: Next time you are b and don’t receive anything.
1: There’s still the show up.
2: That’s a joke.
3: I have also left once only with the show up.
1: Me too.
3: I don’t care about the show up. . .I vote for 0!!!
2: What’s this penny-pinching about?
2: We need a compromise.
1: I still don’t understand your point 2.
3: Yes, democracy! Two against one.
3: Please 2, I have been the sucker too often. I am lucky this time and in the
position to make some money.
3: Transfer = 0?
2: You need those additional 2.5 so desperately?
3: To be honest: yes.
3: . . . and I did not get those 2.5 too often.
1: Why should we give the money away?
3: Why?
3: Quick now, come on, please enter 0?
2: I don’t want to play the upholder of moral standards, but I don’t get your point.
Compromise: 2.
1: I enter 0.
2: I enter 1.5.
1: You can go on being nice after the experiment.
3: Great, we will all get nothing, are you aware of that?
3: You don’t even know to whom you send the money!
Group polarization in the team dictator game reconsidered 39

2: Just like b.
2: 1.5.
1: 0.
3: Apart from that, I am with No. 1: 2:1 for 0. Please accept it No. 2!
2: Hey, 1.5 is obliging.
1: Obliging. . .But completely useless.
3: Please 2, hurry up and say 0.
2: 1.5.
2: Ciao.
3: What are you doing????
1: If we do not hurry up, we all will get nothing.
3: 1, what are you doing?
1: So, what?
3: 2, what’s the deal, that’s impossible!!!!!!!!!!!!
2: 1.5.
1: 1?
2: OK.
3: So, for heavens sake, make that 1!

Individual transfers before team stage/after team stage:

1: 0/0.
2: 2.5/2.5.
3: 2.5/0.

2. The shortest discussion

2: I vote for 1 EURO transfer.

1: That’s OK.
3: For me too.
1: That’s a fast decision.
2: Great.

Individual transfers before team stage/after team stage:

1: 0.5/0.2.
2: 1/2.5.
3: 1.5/0.5.

References

Andreoni, J., & Petrie, R. (2004). Public goods experiments without confidentiality: a glimpse into fund-
raising. Journal of Public Economics, 88, 1605–1623.
Bohnet, I., & Frey, B. S. (1999). Social distance and other-regarding behavior in dictator games: Comment.
American Economic Review, 89, 335–339.
Bornstein, G., & Yaniv, I. (1998). Individual and group behavior in the ultimatum game: are groups more
‘rational’ players? Experimental Economics, 1, 101–108.
40 W.J. Luhan et al.

Bornstein, G., Kugler, T., & Ziegelmeyer, A. (2004). Individual and group decisions in the centipede game:
are groups more ‘rational’ players? Journal of Experimental Social Psychology, 40, 599–605.
Bosman, R., & van Winden, F. (2002). Emotional hazard in a power-to-take experiment. Economic Journal,
112, 147–169.
Camerer, C. F. (2003). Behavioral game theory: experiments in strategic interaction. Princeton: Princeton
University Press.
Cason, T., & Mui, V.-L. (1997). A laboratory study of group polarization in the team dictator game. Eco-
nomic Journal, 107, 1465–1483.
Cooper, D. J., & Kagel, J. H. (2005). Are two heads better than one? Team versus individual play in
signaling games. American Economic Review, 95, 477–509.
Cox, J. C. (2002). Trust, reciprocity, and other-regarding preferences: groups vs. individuals and males
vs. females. In R. Zwick & A. Rapoport (Eds.), Advances in experimental business research (pp.
331–350). Dordrecht: Kluwer Academic Publishers.
Cox, J. C., & Deck, C. A. (2006). When are women more generous than men? Economic Inquiry, 44,
587–598.
Cox, J. C., & Hayne, S. C. (2006). Barking up the right tree: Are small groups rational agents? Experimen-
tal Economics, 9, 209–222.
Davis, D. D., & Harless, D. W. (1996). Group vs. individual performance in a price-searching experiment.
Organizational Behavior and Human Decision Processes, 66, 215–227.
Davis, J. (1992). Some compelling intuitions about group consensus decisions, theoretical and empirical
research and interpersonal aggregation phenomena: Selected examples 1950–1990. Organizational
Behavior and Human Decision Processes, 52, 3–38.
Dufwenberg, M., & Muren, A. (2006). Gender composition in teams. Journal of Economic Behavior and
Organization, 61, 50–54.
Eliaz, K., Ray, D., & Razin, R. (2006). Choice shifts in groups: A decision-theoretic basis. American
Economic Review, 96, 1321–1332.
Fischbacher, U. (2007). z-Tree: Zurich toolbox for readymade economic experiments. Experimental Eco-
nomics, 10, 171–178.
Forsythe, R., Horowitz, J., Savin, N., & Sefton, M. (1994). Fairness in simple bargaining games. Games
and Economic Behavior, 6, 347–369.
Gillet, J., Schram, A., & Sonnemans, J. (2007). The tragedy of the commons revisited: the importance of
group decision-making (Working Paper). University of Amsterdam.
Güth, W., Schmittberger, R., & Schwarze, B. (1982). An experimental analysis of ultimatum bargaining.
Journal of Economic Behavior and Organization, 3, 367–388.
Güth, W., Schmidt, C., & Sutter, M. (2007). Bargaining outside the lab: A newspaper experiment of a
three-person ultimatum game. Economic Journal, 117, 449–469.
Kahnemann, D., Knetsch, J., & Thaler, R. (1986). Fairness as a constraint on profit seeking: Entitlements
in the market. American Economic Review, 76, 728–741.
Kerr, N., MacCoun, R., & Kramer, G. (1996). Bias in judgment: Comparing individuals and groups. Psy-
chological Review, 103, 687–719.
Kocher, M. G., & Sutter, M. (2005). The decision maker matters: individual versus group behavior in
experimental beauty-contest games. Economic Journal, 115, 200–223.
Kocher, M. G., & Sutter, M. (2007). Individual versus group behavior and the role of the decision making
procedure in gift-exchange experiments. Empirica, 34, 63–88.
Kugler, T., Bornstein, G., Kocher, M. G., & Sutter, M. (2007, forthcoming). Trust between individuals
and groups: groups are less trusting than individuals but just as trustworthy. Journal of Economic
Psychology.
Levine, J., & Moreland, R. L. (1998). Small groups. In G. Fisher & G. Lindzey (Eds.), The Handbook of
Social Psychology (Vol. 2, 4th edn., pp. 415–469). Oxford: Oxford University Press.
Maciejovsky, B., & Budescu, D. V. (2007). Collective induction without cooperation? Learning and knowl-
edge transfer in cooperative groups and competitive auctions. Journal of Personality and Social Psy-
chology, 92, 854–870.
Moscovici, S., & Zavalloni, M. (1969). The group as a polarizer of attitudes. Journal of Personality and
Social Psychology, 12, 125–135.
Rockenbach, B., Sadrieh, A., & Mathauschek, B. (2007). Teams take the better risks. Journal of Economic
Behavior and Organization, 63, 412–422.
Stoner, J. (1961). A comparison of individual and group decisions involving risk. Unpublished master
thesis, Massachusetts Institute of Technology, Boston.
Group polarization in the team dictator game reconsidered 41

Sutter, M. (2005). Are four heads better than two? An experimental beauty-contest game with teams of
different size. Economics Letters, 88, 41–46.
Sutter, M. (2007). Are teams prone to myopic loss aversion? An experimental study on individual versus
team investment behavior. Economics Letters, 97, 128–132.
Sutter, M., Kocher, M. G., & Strauß, S. (2006). Individuals and teams in UMTS-license auctions (Working
Paper). University of Innsbruck.
Teger, A., & Pruitt, D. (1967). Components of group risk taking. Journal of Experimental Social Psychol-
ogy, 3, 189–205.
Wildschut, T., Lodewijkx, H. F. M., & Insko, C. A. (2001). Toward a reconciliation of diverging perspec-
tives on interindividual-intergroup discontinuity: the role of procedural interdependence. Journal of
Experimental Social Psychology, 37, 273–285.